Re: [PATCH 3/5] iommu/vt-d: Do deferred attachment in iommu_need_mapping()

[Lu Baolu]

Hi Joerg,

Thanks for doing this.

On 2020/2/18 3:38, Joerg Roedel wrote:
> From: Joerg Roedel <jroedel@xxxxxxx>
>
> The attachment of deferred devices needs to happen before the check
> whether the device is identity mapped or not. Otherwise the check will
> return wrong results, cause warnings boot failures in kdump kernels, like
>
> 	WARNING: CPU: 0 PID: 318 at ../drivers/iommu/intel-iommu.c:592 domain_get_iommu+0x61/0x70
>
> 	[...]
>
> 	 Call Trace:
> 	  __intel_map_single+0x55/0x190
> 	  intel_alloc_coherent+0xac/0x110
> 	  dmam_alloc_attrs+0x50/0xa0
> 	  ahci_port_start+0xfb/0x1f0 [libahci]
> 	  ata_host_start.part.39+0x104/0x1e0 [libata]
>
> With the earlier check the kdump boot succeeds and a crashdump is written.
>
> Signed-off-by: Joerg Roedel <jroedel@xxxxxxx>
> ---
>   drivers/iommu/intel-iommu.c | 6 +++---
>   1 file changed, 3 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/iommu/intel-iommu.c b/drivers/iommu/intel-iommu.c
> index 42cdcce1602e..32f43695a22b 100644
> --- a/drivers/iommu/intel-iommu.c
> +++ b/drivers/iommu/intel-iommu.c
> @@ -2541,9 +2541,6 @@ static void do_deferred_attach(struct device *dev)
>   
>   static struct dmar_domain *deferred_attach_domain(struct device *dev)
>   {
> -	if (unlikely(attach_deferred(dev)))
> -		do_deferred_attach(dev);
> -

This should also be moved to the call place of deferred_attach_domain()
in bounce_map_single().

bounce_map_single() assumes that devices always use DMA domain, so it
doesn't call iommu_need_mapping(). We could do_deferred_attach() there
manually.

Best regards,
baolu

>   	return find_domain(dev);
>   }
>   
> @@ -3595,6 +3592,9 @@ static bool iommu_need_mapping(struct device *dev)
>   	if (iommu_dummy(dev))
>   		return false;
>   
> +	if (unlikely(attach_deferred(dev)))
> +		do_deferred_attach(dev);
> +
>   	ret = identity_mapping(dev);
>   	if (ret) {
>   		u64 dma_mask = *dev->dma_mask;